Identifying the default-mode component in spatial IC analyses of

r Human Brain Mapping 000:000–000 (2011) r

Identifying the Default-Mode Component in SpatialIC Analyses of Patients with Disorders of

Consciousness

Andrea Soddu,1* Audrey Vanhaudenhuyse,1 Mohamed Ali Bahri,2

Marie-Aurelie Bruno,1 Melanie Boly,1,3 Athena Demertzi,1

Jean-Flory Tshibanda,3 Christophe Phillips,2,4 Mario Stanziano,5

Smadar Ovadia-Caro,6 Yuval Nir,7 Pierre Maquet,2,3 Michele Papa,5

Rafael Malach,6 Steven Laureys,1,3 and Quentin Noirhomme1

1Coma Science Group, Cyclotron Research Centre, University of Liege, Liege, Belgium2Cyclotron Research Centre, University of Liege, Liege, Belgium

3Neurology Department, CHU Sart Tilman Hospital, University of Liege, Liege, Belgium4Department of Electrical Engineering and Computer Science, University of Liege, Liege, Belgium

5Medicina Pubblica Clinica e Preventiva, Second University of Naples, Naples, Italy6Department of Neurobiology, Weizmann Institute of Science, Rehovot, Israel7Department of Psychiatry, University of Wisconsin, Madison, Wisconsin

r r

Abstract: Objectives: Recent fMRI studies have shown that it is possible to reliably identify the default-mode network (DMN) in the absence of any task, by resting-state connectivity analyses in healthy vol-unteers. We here aimed to identify the DMN in the challenging patient population of disorders of con-sciousness encountered following coma. Experimental design: A spatial independent componentanalysis-based methodology permitted DMN assessment, decomposing connectivity in all its differentsources either neuronal or artifactual. Three different selection criteria were introduced assessing anti-correlation-corrected connectivity with or without an automatic masking procedure and calculatingconnectivity scores encompassing both spatial and temporal properties. These three methods were vali-dated on 10 healthy controls and applied to an independent group of 8 healthy controls and 11severely brain-damaged patients [locked-in syndrome (n ¼ 2), minimally conscious (n ¼ 1), and vege-tative state (n ¼ 8)]. Principal observations: All vegetative patients showed fewer connections in thedefault-mode areas, when compared with controls, contrary to locked-in patients who showed near-normal connectivity. In the minimally conscious-state patient, only the two selection criteria consider-ing both spatial and temporal properties were able to identify an intact right lateralized BOLDconnectivity pattern, and metabolic PET data suggested its neuronal origin. Conclusions: When assess-

The text reflects solely the views of its authors. The EuropeanCommission is not liable for any use that may be made of the in-formation contained therein.Additional Supporting Information may be found in the onlineversion of this article.

Contract grant sponsor: Israel Science Foundation; Contract grantnumber: 160/07; Contract grant sponsor: MIUR-Progetto FIRBInternazionalizzazione; Contract grant number: RBIN04KW43;Contract grant sponsors: Italian Minerva Grants; European Com-mission (DISCOS-Marie Curie RTN, Mindbridge, DECODER,COST-CATIA), McDonnell Foundation; Mind Science Foundation;

FRS; Reine Elisabeth Medical Foundation and University and Uni-versity Hospital of Liege; European Space Agency.

*Correspondence to: Andrea Soddu, Coma Science Group, Cyclo-tron Research Center and Neurology Department, University ofLiege, Allee du 6 aot n� 8, Sart Tilman B30, 4031 Liege, Belgium.E-mail: [email protected]

Received for publication 7 September 2010; Revised 21 November2010; Accepted 10 December 2010

DOI: 10.1002/hbm.21249Published online in Wiley Online Library (wileyonlinelibrary.com).

VC 2011 Wiley-Liss, Inc.

ing resting-state connectivity in patients with disorders of consciousness, it is important to use amethodology excluding non-neuronal contributions caused by head motion, respiration, and heart rateartifacts encountered in all studied patients. Hum Brain Mapp 00:000–000, 2011. VC 2011 Wiley-Liss, Inc.

Keywords: resting state; fMRI; independent component analysis; default mode; vegetative state;locked-in syndrome; consciousness

r r

INTRODUCTION

Recent progress in fMRI studies on spontaneous brain ac-tivity has demonstrated activity patterns that emerge with-out any task or sensory stimulation, showing promise forthe study of higher order associative network functionality[Biswal et al., 1995; Cordes et al., 2000; Damoiseaux et al.,2006; Fox and Raichle, 2007; Fox et al., 2005; Greicius et al.,2003; Lowe et al., 1998; Mitra et al., 1997; Nir et al., 2006;Vincent et al., 2007; Xiong et al., 1999; Yang et al., 2007]. Oneof the most intensely studied resting-state networks is thedefault-mode network (DMN), which encompasses the pre-cuneus/posterior cingulate (pC), mesiofrontal anterior/ven-tral and posterior parietal cortices [for review, see Fox andRaichle, 2007]. This physiological ‘‘baseline’’ of the humanbrain has been suggested to be related to internally orientedcontent [Vanhaudenuhuyse & Demertzi et al., 2011], self-referential, or social cognition [Schilbach et al., 2008] andhas been referred to as the ‘‘intrinsic network’’ [Fox et al.,2005; Golland et al., 2007]. Interestingly, the ‘‘default’’ net-work can be shown to anticorrelate with an ‘‘extrinsic net-work’’ [Fox et al., 2005; Golland et al., 2008] considered to beimportant in externally oriented tasks [Golland et al., 2007].The potential clinical interest of fMRI studies of DMN con-nectivity is currently being assessed in different pathologiessuch as depression [Anand et al., 2005], schizophrenia [Cal-houn et al., 2009], autism [Kennedy et al., 2006], multiplesclerosis [Lowe et al., 2002], dementia [Greicius et al., 2004;Rombouts et al., 2009], brain death [see case studies by Bolyet al., 2009], and disorders of consciousness [Vanhauden-huyse & Noirhomme et al., 2010]. There are two main waysto analyze resting-state functional connectivity MRI (rs-fcMRI): (1) hypothesis-driven seed-voxel [Fox et al., 2005]and (2) data-driven independent component analysis (ICA)approaches [Kiviniemi et al., 2003; McKeown et al., 1998]—each offering their own advantages and limitations [Coleet al., 2010].

The seed-voxel approach consists in extracting the BOLDtime course from a region of interest (ROI) and determinesthe temporal correlation between this signal (seed) and thetime course from all other brain voxels [Fox et al., 2005]. Tobetter visualize the correlation pattern and better analyze itsproperties, the seed approach can be integrated with graph-theory methods [e.g., Fair et al., 2008; Hagmann et al., 2008].In rs-fcMRI graph representations, the resting-state BOLDtime series for each of the ROIs extracted from the DMN arecorrelated among each other giving a correlation matrix,

which can be represented as a weighted graph. To reducespurious variance unlikely to reflect neuronal activity, theBOLD signal is preprocessed by temporal bandpass filterand spatial smoothing, by regressing out head motioncurves, whole brain signal as well as ventricular and whitematter signal, and each of their first-order derivative terms[Fox et al., 2005]. This straightforward method has beenwidely adopted and offers consistent results [Fox andRaichle, 2007]. However, it has raised some discussionmostly related to the preprocessing procedure, especiallyconcerning the regressing out of the global activity from theBOLD signal, which might induce some spurious anticorre-lations [Fox et al., 2009; Murphy et al., 2009], and the selec-tion itself of the seed regions, which is biased by a prioridefinition.

Contrary to the previous approach, ICA-based rs-fcMRIanalysis [Kiviniemi et al., 2003; McKeown et al., 1998] doesnot require any a priori definition of seed regions. It ana-lyzes the entire BOLD dataset and decomposes it intocomponents that are maximally statistically independent[Hyvarinen et al., 2001]. A number of studies have shownthat ICA is a powerful mathematical tool, which cansimultaneously extract a variety of different coherent neu-ronal networks [De Luca et al., 2006; Esposito et al., 2008;Greicius and Menon, 2004; Greicius et al., 2003, 2004;McKeown et al., 1998] and separate them from other signalmodulations such as those induced by head motion orphysiological confounds [e.g., cardiac pulsation, respira-tory cycle, and slow changes in the depth and rate ofbreathing, Birn et al., 2008; Perlbarg and Marrelec, 2008;Perlbarg et al., 2007]. However, ICA does not provide anyclassification or ordering of the independent components(ICs), and it is usually left to the user to decide which ICcorresponds to the DMN. Automatic approaches havebeen proposed to remove user bias in selecting the compo-nent. For example the ‘‘goodness of fit" is based on match-ing with a previously built template [Greicius et al., 2004].Self-organizing ICA groups components from a group ofsubjects and the user (by visual inspection) selects thegroup that corresponds to the DMN [Esposito et al., 2005].These approaches are based on the spatial extent of thecomponent and some of them have been used in collabora-tion with a power spectrum analysis, removing all compo-nents with more than half of the power due to highfrequencies [for the goodness of fit, see [Vanhaudenhuyse& Noirhomme et al., 2010]. Although having an automaticselection criteria based on spatial properties may be a big

r Soddu et al. r

r 2 r

advantage when the network spatial pattern is still par-tially preserved, it may become a disadvantage for caseswhere the network is mostly destroyed. In these cases,only properly masking the regions of the brain where apreserved activity could be expected would help the selec-tion but then other imaging techniques like, for example,PET should be used at the same time. The possibilityinstead to realize an automatic masking procedure, whichwill be only based on BOLD data, could then be an impor-tant solution for highly pathological cases. At the sametime, giving more importance to time properties otherthan just the power spectrum like temporal entropy orone-lag autocorrelation or to spatial properties no directlypattern-related like, for example, spatial entropy couldalso be a valuable tool to improve the selection power.

The aim of this study is to assess resting-state DMNconnectivity in patients with severe brain damage such asthe vegetative, minimally conscious, or locked-in state, dis-entangling neuronal from artifactual sources. For the clini-cal use of these rs-fcMRI measurements, we first validatedan automated, user-independent spatial ICA-based proce-dure to select and analyze DMN functional integrity inhealthy controls. This approach is based on both spatialand temporal information of the components. The tempo-ral information and spatial properties other than the spa-tial pattern are given by the fingerprint of the component[De Martino et al., 2007]. Using more information enablesa better characterization of artifactual correlations. Threedifferent selection strategies were developed: (1) spatiallypattern driven, (2) based on an automatic masking drivenby the fingerprint properties (time domain dominated),and (3) based on a compromise between spatial and tem-poral properties. We next applied this methodology in anindependent healthy control group and patients in locked-in syndrome [i.e., pseudocoma; LIS, Laureys et al., 2004],minimally conscious state (i.e., showing fluctuating signsof awareness devoid of communication; MCS), or vegeta-tive state (i.e., wakeful unawareness; VS). For the firstselection criterion, which was only spatially patterndriven, we expected to perform equally well in healthycontrols but more poorly than the other two in pathologi-cal brains.

MATERIALS AND METHODS

Subjects and MRI Acquisition

Three independent groups of healthy controls were ana-lyzed for the full study. The first group (Group 1) was an-alyzed for the independent study on DMN ROIs selection(prior knowledge about the DMN equivalent to building aDMN spatial template) and for the creation of an averageDMN fingerprint [De Martino, 2007]. The second group(Group 2) was analyzed to test the DMN selection criteriaintroduced in this manuscript and compare them withother available selection methods (validation of our meth-odology). The third group (Group 3) of healthy controls

was introduced to compare rs-fcMRI analyses results withrespect to patients with disorders of consciousness. Allsubjects went through a resting-state protocol in whichthey were instructed to keep their eyes closed and toremain awake.

Group 1 included 11 healthy volunteers with no neuro-logical or psychiatric history (mean age ¼ 71; SD ¼ 6;range 62–80 years; nine women). Resting-state BOLD datawere acquired on a 3-T MR scanner (Trio Tim, Siemens,Germany) with a gradient echo-planar sequence usingaxial slice orientation (32 slices; voxel size ¼ 3.44 � 3.44 �3.9 mm3; matrix size ¼ 64 � 64 � 32; repetition time ¼2,130 ms, echo time ¼ 40 ms, flip angle ¼ 90�; field ofview ¼ 220 mm). A protocol of 250 scans lasting 533 s wasperformed. A T1-weighted MPRAGE sequence was alsoacquired for registration with functional data on eachsubject.

Group 2 included 10 healthy volunteers with no neuro-logical or psychiatric history (mean age ¼ 21; SD ¼ 3; range18–28 years; four women). Resting-state BOLD data wereacquired on a 3-T MR scanner (Trio Tim, Siemens, Ger-many) with a gradient echo-planar sequence using axialslice orientation (32 slices; voxel size ¼ 3.44 � 3.44 � 3.9mm3; matrix size ¼ 64 � 64 � 32; repetition time ¼ 2,460ms, echo time ¼ 40 ms, flip angle ¼ 90�; field of view ¼ 220mm). A protocol of 350 scans lasting 861 s was performed.A T1-weighted MPRAGE sequence was also acquired forregistration with functional data on each subject.

Group 3 included 8 healthy volunteers with no neuro-logical or psychiatric history (mean age ¼ 48; SD ¼ 13;range 25–65 years; three women). Resting-state BOLD datawere acquired on a 3-T MR scanner (Trio Tim, Siemens,Germany) with a gradient echo-planar sequence usingaxial slice orientation (32 slices; voxel size ¼ 3.0 � 3.0 �3.75 mm3; matrix size ¼ 64 � 64 � 32; repetition time ¼2,000 ms, echo time ¼ 30 ms, flip angle ¼ 78�; field ofview ¼ 192 mm). A protocol of 300 scans lasting 600 s wasperformed. A T1-weighted MPRAGE sequence was alsoacquired for registration with functional data on eachsubject.

Eight patients in VS (mean age ¼ 61; SD ¼ 30; range 16–87 years; all men), one MCS (age 24 years; male), and twoLIS patients (aged 24 years; female and age 20 years; male)were studied with the same scanner used for the thirdgroup of healthy controls. Etiology of the VS was trau-matic in three cases and nontraumatic in five cases; threepatients were studied in the acute (i.e., <1 month postin-jury) and five in the chronic setting. The MCS patient wastraumatic, and the two LIS patients were studied 28months postbasilar artery thrombosis and 4 years aftertrauma (see Table I for patients’ demographic and clinicaldata). All patients underwent repeated behavioral evalua-tions by means of the Coma Recovery Scale Revised [Gia-cino et al., 2004] and the Glasgow-Liege Scale [Born et al.,1985] performed by experienced clinicians (MB, AV, MAB,AD, and SL). The diagnosis was made according to inter-nationally accepted criteria for VS [The Multi-Society Task

r Identifying the DMN in DOC patients r

r 3 r

TABLEI.Patients’demographic,clinical,andstru

cturalim

agingdata

VS1

VS2

VS3

VS4

VS5

VS6

VS7

VS8

MCS

LIS1

LIS2

Gen

der

(age,

years)

Male(62)

Male(21)

Male(16)

Fem

ale(69)

Male(82)

Male(87)

Male(38)

Male(63)

Male(24)

Fem

ale(24)

Male(20)

Etiology

Brainstem

hem

orrhag

eTrauma

Trauma

Anoxia

Vascu

lar

encephalopathy

Trauma

Anoxia

Anoxia

Trauma

Brainstem

stroke

Trauma

Tim

eoffM

RI

(day

safter

insu

lt)

3217

061

561

267

282

2930

185

01,47

5

Outcomeat

12months

Dead

VS

VS

VS

Dead

Dead

VS

MCS

StillMCS

LIS

LIS

Auditory

functiona

None

Startle

reflex

Startle

reflex

Startle

reflex

None

None

Startle

reflex

Startle

reflex

Startle

reflex

Consisten

tmovem

entto

comman

d

Consisten

tmovem

entto

comman

dVisual

functiona

Blinkto

threat

None

None

None

None

None

None

Blinkto

threat

Visual

pursuit

Object

recognition

Object

recognition

Motorfunctiona

Flexionto

pain

Abnorm

alposturing

Abnorm

alposturing

Flexionto

pain

Flexionto

pain

Flexionto

pain

Abnorm

alposturing

Flexionto

pain

Flexionto

pain

Flexionto

pain

Flexionto

pain

Verbal/

Oromotor

functiona

None

Oralreflexes

Vocaliza

tion

Oralreflexes

Oralreflexes

Oralreflexes

Oralreflexes

Vocaliza

tion

Oralreflexes

Oral Movem

ent

Vocaliza

tion

Communicationa

None

None

None

None

None

None

None

None

None

Functional

accu

rate

Functional

accu

rate

Arousala

Without

stim

ulation

Without

stim

ulation

Without

stim

ulation

Withstim

ulation

Withstim

ulation

Withstim

ulation

With stim

ulation

Without

stim

ulation

Without

stim

ulation

Atten

tion

Atten

tion

EEG background

activity

Diffuse

theta-delta,

Right-lateralized

theta-delta,

nonreactive

Irregulardiffuse

theta-delta

Leftlateralized

diffuse

theta-delta

Symmetric

theta-delta,

nonreactive

Irregulardiffuse

theta-delta,

nonreactive

Left-lateralized

theta,

nonreactive

Diffuse

theta-delta,

nonreactive

Notav

ailable

Diffuse

theta,

tran

sien

tdelta

Notav

ailable

Lesionson

structural

MRI

Pontine

hem

orrhag

eextendingto

midbrain

Diffuse

axonal

injury.Bilateral

subduralfrontal

hygroma

Diffuse

atrophy

withsecondary

hydrocephalus.

Bilateral

lesions

oflenticu

lar

nucleu

sthalam

us,

and

parah

ippocampal

structures

Diffuse

bilateral

whitematter

atrophy,caudate,

thalam

us,

and

parah

ippocampal

lesions

Diffuse

cortical

andsu

bcortical

atrophy.Bilateral

basal

gan

gliaan

dwhite

matterlesions

Subarachnoid

hem

orrhag

e,brainstem

,left

lenticu

larnucleu

s,temporalan

dmultifocal

frontoparietal

contusions

Diffuse

white

matterlesions

Posterior

occipitoparietal

isch

emia

Corpuscallosu

m,

leftthalam

us,

pontine,

bilateral

frontoparietallesions

Cen

tro-

protuberen

tial

lesion

Pontine,

midbrain,

cerebellum,

leftthalam

us

lesions

VS,veg

etativestate;

LIS,locked

-insyndrome;

MCS,minim

ally

consciousstate.

aBased

onComaRecoveryScore-Rev

ised

assessmen

t.

Force on PVS, 1994], MCS [Giacino et al., 2004], and LIS[American Congress of Rehabilitation Medicine, 1995].

Written informed consent was obtained from all healthyvolunteers and from the legal representative of all patients(and from the LIS patients using eye-coded communica-tion). The study was approved by the Ethics Committee ofthe Faculty of Medicine of the University of Liege, whichcomplies with the Code of Ethics of the World MedicalAssociation (Declaration of Helsinki).

Data Preprocessing and Analysis

fMRI data were preprocessed using the ‘‘BrainVoyager’’software package (R. Goebel, Brain Innovation, Maastricht,The Netherlands). Preprocessing of functional scansincluded 3D motion correction, linear trend removal, slicescan time correction, and filtering out low frequencies ofup to 0.005 Hz. The data were spatially smoothed with aGaussian filter of full width at half maximum value of 8mm. The functional images from each subject were alignedto the participant’s own anatomical scan and warped intothe standard anatomical space of Talairach and Tournoux[1988]. The Talairach transformation was performed intwo steps. The first step consisted in rotating the 3D dataset of each subject to be aligned with stereotaxic axes (forthis step, the location of the anterior commissure (AC), theposterior commissure (PC), and two rotation parametersfor midsagittal alignment was specified manually). In thesecond step, the extreme points of the cerebrum werespecified. These points together with the AC and PC coor-dinates were then used to scale the 3D data sets into thedimensions of the standard brain of the Talairach andTournoux [1988] atlas using a piecewise affine and contin-uous transformation. ICA [Formisano et al., 2004; Hyvari-nen et al., 2001] was performed with the ‘‘BrainVoyager’’software package using 30 components [Jafri et al., 2008;Koch et al., 2010; Meindl et al., 2010; Weissman-Fogelet al., 2010; Ylipaavalniemi and Vigario, 2008]. A spatialmean was subtracted from all the voxels (mean calculatedall over the voxels at a fixed time), and principal compo-nent analysis was applied for dimensionality reductionbefore performing ICA.

Decomposing Connectivity in Independent

Component Graphs

We developed here a new methodology, which takesadvantage of the capability of ICA, to decompose the sig-nal in neuronal and artifactual sources but still preservesthe concept of connectivity in a defined network of ROIs.This methodology in fact allows building for each IC aconnectivity graph, which summarizes the level of connec-tivity for a defined network of ROIs according to the timebehavior described by the correspondent IC time course.Our connectivity study used 13 ROIs defined on an aver-age DMN map calculated on a group of 11 healthy sub-

jects (Group 1). We performed, as implemented in BrainVoyager [self-organizing ICA, Esposito et al., 2005], a spa-tial similarity test on single-subjects ICs and we averagedthe maps belonging to the cluster, which was selected byvisual inspection as DMN. The selected 13 ROIs were themost representative regions of our average DMN mapclose to the DMN target points previously published [Fairet al., 2008; Fox et al., 2005] including: medial prefrontalcortex ventral (MFv) [�3, 39, �2], medial prefrontal cortexanterior (MFa) [2, 59, 16], pC [�3, �55, 21], left posteriorparietal lobe (L-pP) [�49, �60, 23], right posterior parietallobe (R-pP) [45, �61, 21], left superior frontal gyrus (L-sF)[�19, 32, 51], right superior frontal gyrus (R-sF) [23, 29,51], left middle temporal gyrus anterior [�61, �11, �10],right middle temporal gyrus anterior [57, �11, �13], leftparahippocampal/mesiotemporal [�23, �17, �17], rightparahippocampal/mesiotemporal [25, �16, �15], left thala-mus (L-T) [�5, �11, 7], and right thalamus (R-T) [4, �11,6] (the ROIs were set initially to a cubic shape 10 � 10 �10 mm3, and the center was chosen accordingly to theDMN extracted from Group 1, but once the ROI wassaved in Brain Voyager only the ROI’s voxels belonging tothe DMN end up making the saved ROI).

To study connectivity between each pair of target points,we implemented a new method, which is based on ICAfollowed by a general linear model (GLM). After runningICA with 30 components, we used the corresponding timecourses to regress in the BOLD signal in each of the 13ROIs. The time courses from each ROI were extracted asthe arithmetic mean of the time courses of the voxelsbelonging to the same ROI. Note that by using a smooth-ing of a 8-mm kernel implies that taking an arithmeticmean as implemented in Brain Voyager or extracting thefirst eigenvariate will give similar results. For each compo-nent, we obtained 13 parameter estimates (beta values)indicating the weight of each regressor and the corre-sponding T-values (see Appendix). To build a connectivitygraph, we drew an edge between each pair of target pointswith T > Tth with Tth corresponding to 1 � P/78 for P ¼0.05 with 267 degrees of freedom (Bonferroni correctionfor multiple comparisons was performed dividing P bythe number of possible edges between the 13 nodes; 13*(13� 1)/2 ¼ 78). To account for the fact that ICA does notpredict the sign of the ICs, the condition T < �Tth wasalso used. This allowed us to end up with two connectiv-ity graphs for each of the 30 components (1–30 for the con-dition T > Tth and 31–60 for T < �Tth). We hypothesizedthat the number of edges E for each of the 60 connectivitygraphs should be the highest for the component of theDMN. However, given that no regressing out of the globalsignal was applied, we did not pick the component corre-sponding to the graph with the largest number of totaledges (i.e., the global component could appear as the mainsource of connectivity). Therefore, we implemented ‘‘anti-correlation-corrected number of edges’’ (see Fig. 1). Theanticorrelation-corrected number of edges was obtained bymultiplying the total number of edges of each graph by a


r 5 r

Figure 1.

Illustration of the default-mode selection method in a minimally con-

scious-state patient. (a) Selection of the IC corresponding to the

graph with the highest number of edges. (b) Selection of the IC cor-

responding to the graph with the highest number of global edges (to

select the global signal IC). (c) Selection of the IC corresponding to

the graph with the highest number of anticorrelation-corrected

edges. (d) Selection of the IC corresponding to the graph with the

highest anticorrelation-corrected score. Right panel: (a) number of

edges, (b) number of global edges, (c) number of anticorrelation-cor-

rected edges, and (d) anticorrelation-corrected score of each graph

vs. the corresponding IC number. Middle panel: Spatial map of the

selected IC. Left panel: Connectivity graph of the selected IC. MFv,

medial prefrontal cortex ventral; MFa, medial prefrontal cortex ante-

rior; pC, posterior cingulate/precuneus; pP, posterior parietal lobe;

sF, superior frontal gyrus; aT, middle temporal gyrus anterior; mT,

parahippocampal/mesiotemporal; T, thalamus. Left is right side of

brain.

r Soddu et al. r

r 6 r

weight ‘‘w,’’ which measures the anticorrelation of theDMN activity with the extrinsic system/external controlnetwork (ECN). Similarly to the 13 targets ROIs of theDMN, we selected as ROIs the five most representativeregions of the extrinsic system appearing as anticorrelatedregions in the DMN average map calculated on the Group1 of healthy subjects close to the ECN target points previ-ously published (Fox et al., 2005] including: (left supra-marginal gyrus [�56, �33, 37], right supramarginal gyrus[54, �39, 38], left middle temporal gyrus posterior [�52,�53, �5], right middle temporal gyrus posterior [52, �57,�5], and supplementary motor area [2, 5, 46]). The anticor-relation-corrected number of edges EAntiCC was thendefined in terms of the number of edges E as EAntiCC ¼E*w (see Appendix), with the anticorrelation index w a

number between 0 and 1. The index would be close to 0 ifthe activity in the extrinsic network highly correlates withthe activity in the intrinsic network, e.g., a global signal,and close to 1 if the activity in the extrinsic network isanticorrelated with the activity in the DMN, e.g., the DMNcomponent. A new number of edges could be built by justinverting the � signs in the definition of EAntiCC (see Ap-pendix) obtaining a new number, which we called ‘‘globaledges’’ and which is highest for the global component (seeFig. 1). This allowed us to isolate the global mode and toremove it from the full set of components.

Selection Criteria

Three different strategies were developed: (1) spatiallypattern driven, (2) based on an automatic masking driven

Figure 2.

Flow chart illustrating the second selection criterion. Arrow indicates the starting point in the

flow chart. Different color boxes indicate the different steps. [Color figure can be viewed in the

online issue, which is available at wileyonlinelibrary.com.]


r 7 r

by the fingerprint properties (time domain dominated),and (3) based on a compromise between spatial and tem-poral properties.

First selection criterion

This selection criterion selects as the DMN the componentwith the highest number of anticorrelation-corrected edges(see Fig. 1). The number of anticorrelation-corrected edgesof the component selected is indicated in Tables I and II ofthe Supporting Information with the name ‘‘EAntiCC init.’’

Second selection criterion

This selection criterion is still based on the highest num-ber of edges but after performing an automatic maskingobtained by removing subsequently ROIs from the net-work (see flow chart in Fig. 2 and Appendix for the defini-tion of the distance DIC). The automatic maskingproceeded by removing subsequently (going through allthe 13 ROIs) one ROI (first step), then two regions together(second step), and three regions up to five regions (laststep). At each step, the component selected as the onewith the highest number of anticorrelation-corrected edgesin the reduced network and with the minimum distanceamong all the possible reduced networks (for that step)was tested by comparing its fingerprint (as in De Martinoet al. [2007], the fingerprint reports normalized values,respectively, for: degree of clustering, skewness, kurtosis,spatial entropy as calculated from the distribution of the zvalues for the considered IC and one-lag autocorrelation,temporal entropy, and the power of the five frequencybands [0–0.008, 0.008–0.02, 0.02–0.05, 0.05–0.1, and 0.1–0.25Hz] as extracted from the time course of the consideredIC) with an average fingerprint (distance DIC from the ref-erence fingerprint needed to be inside two standard devia-tions with the standard deviation calculated on eachsubject IC distances’ distribution) built from the DMNcomponents of the 11 healthy subjects (Group 1). Thenumber of anticorrelation-corrected edges of the compo-nent selected is indicated in Tables I and II Supporting In-formation with the name ‘‘EAntiCC fin.’’

Third selection criterion

The third automatic selection criterion selects the com-ponent with the highest ‘‘anticorrelation-corrected score’’(SAntiCC), built by multiplying the number of anticorrela-tion-corrected edges by a new weight ‘‘wF,’’ which meas-ures the distance of its fingerprint from the averagefingerprint of the DMN in healthy controls (Group 1). Theweight wF is close to 0 for components that have ‘‘artifac-tual’’ source and close to 1 for components with ‘‘neuro-nal’’ origin—the latter assumes that in healthy controlsICA was able to fully separate artifactual from neuronalsources. The number of anticorrelation-corrected edges ofthe component selected is indicated in Tables I and II

of the Supporting Information with the name ‘‘EAntiCC

(SAntiCC max).’’The three selection criteria were compared with self-

organizing ICA, which selected for each single subject themost spatially similar component to a DMN average mapbased on the 11 healthy subjects of Group 1 (self-organiz-ing ICA used to cluster 30 average maps vs. 30 ICs maps).

Single-Subject and Group Analysis

For each single subject, we presented a connectivitygraph showing thin and thick lines in three colors: red, or-ange, and yellow. Thin and thick lines correspond, respec-tively, to the conditions T > Tth and T*w > Tth [thin linefor edges and thick line for ‘‘weighted edges’’ (EW)], andthe three colors: red, orange, and yellow correspond,respectively, to P values P ¼ 0.05, P ¼ 0.01, and P ¼ 0.001.It is important to note that the number of anticorrelation-corrected edges, as previously stated, is calculated by thenumber of edges E (number of edges in the graph corre-sponding to the condition T > Tth for P ¼ 0.05 multipliedby the corresponding anticorrelation index w) and is notobtained by counting the number of edges correspondingto the conditions T*w > Tth termed ‘‘weighted edges" (seeTables I and II in the Supporting Information to comparevalues).

Two-tailed unequal variance Student’s t-test comparedthe total number of edges, the anticorrelation index, theanticorrelation-corrected total number of edges, the totalnumber of weighted edges, the anticorrelation-correctedscores, and the ratio of the anticorrelation-corrected scoreover the anticorrelation-corrected total number of edges inthe DMN in controls vs. VS patients for each of the threedifferent selection criteria.

Spatial maps were obtained by running a two-step anal-ysis. First, the BOLD signal from each single subject wasregressed with the time courses of the ICs selected asDMN as selected by the second selection criterion (auto-matic masking). Second, the estimated parametric mapsentered a multisubject random-effect analysis providinggroup-level statistical T-maps. Maps were thresholded at afalse discovery rate corrected P < 0.05 within DMN maskobtained from an independent dataset (Group 1) shown asblack and white contour volume of interest in Figures 3and 5–7. The mean connectivity graph was derived bydrawing an edge between each pair of ROIs with a sum ofanticorrelation-corrected beta values beta*w significantlydifferent (permutation test) from the absolute value oftheir difference. Different colors correspond to differencesin significance, with red, orange, and yellow representingP ¼ 0.05, P ¼ 0.01, and P ¼ 0.001, respectively. Thickerlines are the connections that survive a multicomparisonBonferroni correction with the number of possible edgesbetween the 13 nodes: 13*(13 � 1)/2 ¼ 78.

A contrast T-test map indicates the regions where theDMN time courses for controls and VS patients predicted

r Soddu et al. r

r 8 r

in a significantly different way the BOLD signal withrespect to the mean signal. The difference graph shows aconnection between each pair of ROIs for which the sumof anticorrelation-corrected beta values minus their differ-ence is significantly different (permutation test) betweencontrols and VS patients. The colors and thickness for theedges in the difference graph have the same meaning as inthe mean graph.

RESULTS

Healthy Controls

In Group 2 of controls (see Table I in the Supporting In-formation), the three different connectivity ICA methodsidentified the same DMN at the single-subject level. Crossvalidation with self-organizing ICA showed concordanceof the selected network in all healthy volunteers. In Group3 of controls (see Table II in the Supporting Information),cross validation with self-organizing ICA showed concord-ance of the selected network in six of eight healthy volun-teers. The time course of the selected DMN showed a

power spectrum peaking in the range of 0.02–0.05 Hz. Atthe group level (Fig. 4), the nodes showing the highestnumber of significant connections (edges surviving Bonfer-roni correction) were the ventral and anterior medial pre-frontal cortices (number of edges ¼ 7) followed by theprecuneus/PCC (number of edges ¼ 4), and left and rightposterior parietal cortices (number of edges ¼ 3). Healthycontrols’ mean spatial map identified the DMN pattern asshown in Figure 3. As confirmed by the connectivitygraph, the regions that are the predominant representa-tives of the DMN are MFa and MFv, pC, L-pP and R-pP,and L-sF and R-sF.

VS Patients’ Group Data

VS patients’ spatial map (Fig. 3) showed no significantpattern as also confirmed by the connectivity graph (timecourses corresponding to the DMN selected by the secondselection criterion have been used). The contrast controlvs. VS patients’ spatial map did not show any significantresult when thresholded at false discovery rate correctedP < 0.05 within a DMN mask obtained from an

Figure 3.

Upper part: Random-effect group analyses identifying the

default-mode network (DMN) in eight healthy controls and eight

patients in a vegetative state (VS). Results are thresholded at

false discovery rate corrected P < 0.05 with a mask given by

the black and white contour regions showing the DMN from an

independent dataset (n ¼ 11 healthy controls, Group 1). Lower

part: Graphical representation (i.e., fingerprint; normalized val-

ues) of DMN temporal properties (five frequency bands, tempo-

ral entropy, and one-lag autocorrelation) and spatial properties

(spatial entropy, skewness, kurtosis, and clustering) in healthy

controls [mean (yellow) and SD (green)] and in VS patients

[mean (cyan) and SD (blue)]. [Color figure can be viewed in the



r 9 r

independent dataset of healthy controls (Group 1). On thecontrary, differences in connectivity graphs highlightedthe connection between the MFv and the pC as well asbetween the MFv and the left and right lateral parietal, theL-sF and R-sF, the left anterior temporal gyrus, and theMFa. Compared with controls, VS patients’ graphs showeda significantly lower total number of edges (numbers areapproximated to integer, 41 � 13, range 21–66 vs., respec-tively, for the three selection criteria: first: 17 � 5, range15–28, P < 0.001; second: 10 � 4, range 6–15, P < 0.001;third: 14 � 2, range 10–15, P < 0.001) as well as the anti-correlation index (0.72 � 0.09, range 0.55–0.83 vs., respec-tively, for the three selection criteria: first: 0.57 � 0.08,range 0.42–0.69, P ¼ 0.004; second: 0.52 � 0.10, range 0.41–0.65, P < 0.001; third: 0.56 � 0.08, range 0.43–0.69, P ¼0.003). Significant differences (which confirmed previousresults) were observed in the total number of anticorrela-tion-corrected edges (numbers are approximated to inte-ger, 30 � 12, range 14–53 vs., respectively, for the threeselection criteria: first: 10 � 2, range 8–12, P ¼ 0.002; sec-ond: 3 � 2, range 3–9, P < 0.001; third: 8 � 2, range 4–10,P < 0.001) as well as total number of weighted edges (18� 10, range 4–29 vs., respectively, for the three selectioncriteria: first: 1 � 2, range 0–6, P ¼ 0.002; second: 1 � 1,range 0–3, P ¼ 0.002; third: 1 � 2, range 0–6, P ¼ 0.002).For the anticorrelation-corrected scores, which summar-

ized in a single number both spatial and temporal proper-ties, compared with controls, VS patients showed a loweranticorrelation-corrected score (numbers are approximatedto integers, 21 � 9, range 7–38 vs., respectively, for thethree selection criteria: first: 2 � 1, range 1–4, P < 0.001;second: 3 � 1, range 1–5, P < 0.001; third: 3 � 1, range 1–5, P < 0.001). The weight wF, which is a measure of the na-ture of the IC selected [artifactual (low) vs. neuronal(high)], showed, compared with controls, lower values inVS patients (0.68 � 0.14, range 0.47–0.84 vs., respectively,for the three selection criteria: first: 0.24 � 0.07, range0.16–0.36, P < 0.001; second: 0.48 � 0.18, range 0.24–0.74, P¼ 0.03; third: 0.40 � 0.19, range 0.16–0.70, P ¼ 0.005). InFigure 3, the mean DMN fingerprint for controls (yellowline) showed the characteristic bird shape [De Martinoet al., 2007] with a predominant frequency band of 0.02–0.05 Hz, high level of clustering, and one-lag autocorrela-tion. The mean DMN fingerprint (DMN components wereselected with the second selection criterion) for VS patients(cyan line) did not show a normal bird-like shape butshowed a predominant frequency of 0.1–0.25 Hz even ifthe contribution in the frequency band of 0.02–0.05 Hzwas still important. Significant lower temporal entropywas also observed in VS patients with respect to healthycontrols. Finally, VS patients moved in the scanner morethan healthy volunteers showing compared with controls

Figure 4.

Connectivity graphs for the eight healthy controls and the eight VS patients’ groups and between-

group differences. Red (blue), orange (cyan), and yellow (green) lines represent P ¼ 0.05, P ¼ 0.01,

and P ¼ 0.001, respectively, for positive and negative differences. Thicker lines are connections sur-

viving correction for multiple comparisons. Nodes are defined as for Figure 1. [Color figure can be

viewed in the online issue, which is available at wileyonlinelibrary.com.]

r Soddu et al. r

r 10 r

Figure 5.

Single-subject analyses identifying the default-mode network

(DMN) and connectivity graphs in (a) a representative healthy

control, (b, c) LIS patients 1, 2, and (d) MCS patient. Positive cor-

relations (yellow) and anticorrelations (blue) with the DM time

course shown on a transverse section at Z ¼ 24 mm (thresholded

at corrected P < 0.05). Black and white contour regions show the

DMN from an independent dataset of 11 healthy controls. Motion

curves illustrate translation (in mm) for x (red), y (green), and z

(blue) and rotation (in �) for pitch (yellow), roll (purple), and yaw

(cyan) parameters, and the DMN time course illustrates the nor-

malized BOLD signal over 600 s. The fingerprint summarizes the

DMN temporal and spatial properties for each subject (red) super-

imposed to the control data of eight healthy subjects (mean in yel-

low and standard deviation in green). The connectivity graph

illustrates the connections between the 13 selected DM nodes at

different thresholds for significance (thick lines, ‘‘weighted edges’’

are corrected for external network anticorrelations). Nodes are

defined as for Figure 1.


r 11 r

higher even if not significantly different ‘‘displacement’’and ‘‘speed’’ (see Appendix for definition; displacement ¼0.95 � 0.67, range 0.2–2.2 vs. 2.3 � 2.0, range 0.2–5.9, P ¼0.11; speed ¼ 0.07 � 0.04, range 0.02–0.13 vs. 0.28 � 0.30,range 0.08–0.96, P ¼ 0.08).

Patients’ Single-Subject Analysis

Single-subject analysis refers to the DMN selected usingthe second criterion. None of the VS patients showed aDMN with both the spatial and temporal patterns compa-rable with healthy controls see Fig. 5 for an example ofhealthy control and Figs. 6 and 7 for the VS patients.Rapid transient ‘‘clonic’’ movements were observed for allpatients. Spatial maps showed periphery patterns charac-teristic of motion artifacts. Individual fingerprint analysisshowed a shift toward the higher frequency bands (0.1–0.25 Hz) especially in VS patients 1, 4, 6, and 8. In patients1 and 8, a contribution in the low frequency band (0.02–0.05 Hz) was also observed, whereas for patients 2, 3, 5,and 7, the component selected was dominated by a lowfrequency behavior (0.02–0.05 Hz). Graph analysis showedsome residual connections (weighted edges) in patient 2(between pC and right lateral parietal posterior and ante-rior temporal gyri), in patient 5 (between L-sF and R-sF),and in patient 8 (between pC and right lateral parietalposterior and L-T). In patient 2, we also observed substan-tial movement throughout the recording with a highlycorrelated power spectrum of the BOLD time course andthe motion parameter curves.

The two locked-in patients (Fig. 5) showed near-normalDMN connectivity spatial pattern despite substantialmovement artifacts (the three selection criteria gave thesame component). In the temporal domain, an increase inthe high frequencies was observed for patient 1 even if thecontribution in the frequency band (0.02–0.05 Hz) was pre-served. For patient 2, a proper fingerprint was observed.The connectivity graphs were characterized, respectively,by 28 and 36 total number of edges and an anticorrelationindex of 0.74 for both patients giving, respectively, 21 and27 total number of anticorrelation-corrected connections,which were not significantly different from healthy con-trols (P ¼ 0.27) and counted 21 and 28 weighted edges(P ¼ 0.25). The anticorrelation-corrected scores were,respectively, 14.0 and 16.6 in line with healthy controls.The weight wF with values 0.67 and 0.62 was also in thehealthy volunteers’ range. The MCS patient showed aproper DMN spatial pattern in the right hemisphere asalso confirmed by the connectivity graph (Fig. 5). A propertime-course behavior was also observed together with aproper fingerprint. The total number of edges was 10 (sec-ond and third selection criteria) with an anticorrelationindex of 0.65 giving a number of anticorrelation-correctededges of 6.5, the weighted edges 6, and a score of 3.6 withthe anticorrelation-corrected edges and the anticorrelation-corrected score consistent with VS patients, but with theweighted edges being lower of healthy controls and higher

than VS patients. The weight wF was consistent withhealthy controls.

DISCUSSION

Here, we have assessed different methodologies aimingto identify the DMN based on spatial ICA in the challeng-ing patient population of severe brain damage. Three dif-ferent strategies were developed: (i) spatially patterndriven, (ii) based on an automatic masking driven by thefingerprint properties (time domain dominated), and (iii)based on a compromise between spatial and temporalproperties. After testing that the three strategies selectedthe same DMN component on an independent group of 10healthy controls, we confirmed the validity of the threestrategies by comparing results with those obtained byrunning self-organizing ICA as implemented in Brain Voy-ager (similar results were obtained by preprocessing datain SPM and running ICA in FSL and then comparing ourselection criteria with the similarity or the goodness of fittests with or without selecting components based on theirpower spectrum). We next applied our methodology toanother independent group of 8 healthy controls, eight VS,two LIS, and one MCS patients. Each of the three connec-tivity spatial ICA strategies permitted to identify the DMNat the single-subject level in a user-independent manner.In the group of 8 healthy controls, the three methods iden-tified the same component characterized by the typicalspatial pattern of the DMN [Beckmann et al., 2005] accom-panied by a corresponding time course with a power spec-trum peaking in the range 0.02–0.05 Hz. Cross validationwith self-organizing ICA showed concordance of theselected network in six of eight healthy volunteers. In thetwo subjects where both methods disagreed, post hoc anal-ysis showed that self-organizing ICA identified a networkwith a spatial pattern resembling the DMN but with atime course peaking at higher frequency with respect tothe time course of the network selected by our methodol-ogy, which was dominated by neuronal frequencies [e.g.,0.01–0.05 Hz, De Luca et al., 2006]. The DMN fingerprintconfirmed the previously reported characteristic bird-shaped pattern in all healthy controls with a predominantfrequency band of 0.02–0.05 Hz, high level of clustering,and one-lag autocorrelation [De Martino et al., 2007].

For VS patients, the three selection criteria did not agreeon the selected component. To have an indication of whichof the three criteria, which worked equally well in healthysubjects, could be the one with more selection power inhighly pathological brains, we decided to test our threemethods on a MCS patient. In this patient, the DMN waseasily detectable by visual inspection of the spatial patterntogether with the fingerprint. The presence of the DMNonly in the right hemisphere was in accordance with cere-bral metabolic PET data, suggesting a neuronal activitypattern consistent with healthy controls only in the righthemisphere. The finding that only the second and thethird selection criteria were able to pick the right

r Soddu et al. r

r 12 r

component as DMN in this patient may illustrate theadvantage of these last two methods in highly pathologicalbrains where spatial patterns may be partially destroyed.To detect residual patterns of neuronal activity, it is prob-ably more important to give more or equal selecting powerto time than to space. Given the obtained results, we

decided to use the second selection criterion because itsfingerprint-based procedure guiding the automatic mask-ing seemed promising to identify for the DMN a compo-nent with neuronal source. Using our second selectioncriterion, our method did not identify a single connectionbetween DMN nodes, which reached statistical

Figure 6.

Single-subject analyses identifying the default-mode network (DMN) and connectivity graphs in

(a–d) VS patients 1–4. See Figure 5 for explanatory legend. [Color figure can be viewed in the



r 13 r

significance at the VS patient’ group level and failed toshow any significant voxel that showed a consistent pat-tern of DMN connectivity. The mean DMN fingerprintwas not bird shape like (even if the second selection crite-rion gave the most similar fingerprint to healthy controls)but showed a predominant frequency of 0.1–0.25 Hz,

which was significantly higher from that observed inhealthy controls. It is important to stress that broadbandultra-slow fluctuations (<0.1 Hz) in BOLD fMRI and inneuronal activity typically show a 1/f power profile [Niret al., 2008] with no predominant frequency, whereas nar-row-band spontaneous fluctuations around 0.1 Hz (as

Figure 7.

Single-subject analyses identifying the default-mode network (DMN) and connectivity graphs in

(a–d) VS patients 5–8. See Figure 5 for explanatory legend. [Color figure can be viewed in the


r Soddu et al. r

r 14 r

observed in the VS group) have been shown to be impli-cated in non-neuronal processes such as vasomotion[Mitra et al., 1997]. This shift toward higher frequencies inthe DMN could be caused by pulse and respiration arti-facts (and their aliasing) [De Luca et al., 2006]. With a TRof 2 s (corresponding to a sampling rate of 0.5 Hz and aNiquist frequency (Nf) of 0.25 Hz) aliasing may be an im-portant source of artifact. For example, a respiratory rateof 12 breaths per minute (i.e., 0.2 Hz < Nf) will not bealiased, while its first harmonic of 0.4 Hz (i.e., >Nf) isaliased at |0.4–0.5 Hz| ¼ 0.1 Hz. Similarly, a heart rate of(for example) 63 beats per minute (1.05 Hz) is aliased at|1.05–2*0.5 Hz| ¼ 0.05 Hz appearing in the DMN charac-teristic frequency range. The frequency band that waslower in VS patients when compared with controls was0.008–0.02 Hz possibly reflecting a decrease in DMN neu-ronal activity. The temporal entropy and one-lag autocor-relation were also shown to be decreased in VS patientspossibly explained by, respectively, movement artifactsand high-frequency time-course behavior caused by pulseand respiratory artifacts. The clustering, skewness, kurtosis,and spatial entropy were not different between VS patientsand controls, presumably indicating that the spatial proper-ties of neuronal DMN activity and artifacts (i.e., pulse, respi-ration, movement, and global effects) might be comparable.However, this does not imply that artifacts would show asimilar spatial pattern as the DMN. When the DMN doesnot exist or some other component (e.g., motion, heart, or re-spiratory rate) gives a higher contribution to connectivity—even if a fingerprint-driven selection criterion tends to sup-press this scenario–then the artifactual IC provides an upperbound on the connectivity number of edges for the DMN(i.e., a component is de facto always selected).

The connectivity study showed no overlap between thenumber of ‘‘anticorrelation-corrected connections’’ or edgesin VS patients and in controls. Connectivity graph analysisidentified anterior–posterior midline disconnections in VS,in line with previous studies emphasizing the critical roleof the precuneus/PCC and mesiofrontal cortices in theemergence of conscious awareness [Boly et al., 2008; Van-haudenhuyse & Noirhomme et al., 2010]. Indeed, theseareas are among the most active brain regions in consciouswaking [Gusnard et al., 2001] and are among the leastmetabolically active regions in VS [Laureys et al., 2004]and in other states of altered consciousness such as gen-eral anesthesia [Alkire and Miller, 2005], sleep [Maquet,1997], hypnotic state [Rainville et al., 2002], or dementia[Minoshima et al., 1997]. It has been suggested that theserichly connected multimodal medial associative areas arepart of the neural network subserving consciousness[Baars et al., 2003] and self-awareness [Hagmann et al.,2008; Laureys et al., 2007]. It is important to stress herethat the difference in anticorrelation-corrected connectivitymeasured between VS patients and healthy controls is acombined result of both the reduction in the positivecorrelation between the regions of the DMN and of theanticorrelation with the regions of the ECN. A reduction

is in fact observed in both the number of edges and theanticorrelation index in VS patients with respect to healthycontrols (see Table II in the Supporting Information).

At the single-subject level, none of the VS patientsshowed a DMN with a spatial and temporal pattern thatwas comparable with that observed in healthy controls. Itshould be noted that only in patient VS2 as for the MCSpatient the metabolic PET study reported a preserved neu-ronal activity only in the right hemisphere, which was thehemisphere showing a preserved DMN spatial pattern (asillustrated by the connectivity graph) and with a DMNtime course consistent with neuronal activity. Individualfingerprint analysis showed a shift toward the higher fre-quency bands (0.1–0.25 Hz) in four of eight VS patients(VS cases 1, 4, 6, and 8; see Figs. 6a,d and 7b,d) possiblycaused by pulse or respiration artifacts (or by their alias-ing). Aliasing of pulse and respiration frequencies couldalso explain the increase in lower frequency bandsobserved in VS patients (VS 2, 3, 5, and 7) where normalDMN neural activity is known to peak (see Figs. 6b,c and7a,c). However, in the absence of simultaneous recordingof heart and respiratory rates during the fMRI acquisi-tions, this remains speculative and one cannot here for-mally exclude the presence of a neuronal contribution.Note that in addition to simultaneous recording of physio-logical parameters, future studies should use a shorter TR,which will strongly reduce the aliasing of high frequency(albeit covering a fewer number of slices). Graph analysesalso showed some residual connections (corrected forglobal effects) in VS patient 2 (between pC and right lat-eral parietal posterior and right anterior temporal gyrus),in VS5 (between R-sF and L-sF), and VS8 (between pC andright lateral parietal posterior and L-T). In our view, theseresults may not reflect residual DMN neuronal activity butcan be explained by pulse, respiratory, and mainly move-ment artifacts. The particular case of VS2 deserves somemore caution because of the consistency between the spatialpattern and the independent observed metabolic PET data(also clinically the patient was considered a borderline MCScase). Patient VS8 was the only VS patient that subsequentlyrecovered minimal signs of consciousness.

Despite substantial movement artifacts, the two locked-in patients showed in the spatial domain a near-to-normalDMN connectivity pattern. In the temporal domain, themovements caused for LIS1 an increase in the high fre-quency band without masking the neuronal contributionin the lower frequency bands. The graph analysis showeda connectivity pattern indistinguishable from that observedin healthy controls. The MCS patient showed a DMN pat-tern intermediate between healthy controls and VSpatients, suggesting that the extracted variables used forindividual graph analyses may offer a classification of thelevel of consciousness in these challenging patients—ashas been shown at the group level [Vanhaudenhuyse &Noirhomme et al., 2010]. When dealing with DOCpatients, major confounds in resting-state fMRI acquisi-tions are movement, pulse, respiration (and their aliasing),


r 15 r

and global-effect artifacts [Soddu et al., 2011]. Future clini-cal studies should therefore perform simultaneous heartand respiratory rate recordings [Gray et al., 2009] and realmovement monitoring.

Normalization procedures are also very important issueswhen dealing with severe brain injury and highlydeformed or, especially for chronic cases, atrophied brains[Dai et al., 2008]. In this study, anatomical placements ofall ROI positions were visually checked to make sure noartifactual signal from non-neuronal ventricular or whitematter structures was included. Finally, it is important tostress that our presented methodology cannot protectagainst the well-known problem of oversplitting or under-splitting of brain networks, inherent to ICA. In the case ofICA decompositions, higher dimensionalities of the modelhave recently been advocated [Kiviniemi et al., 2009; Smithet al., 2009], although the robustness of a given level ofdecomposition relies on being supported by data quality.The highly pathological brains of DOC patients do notallow quite often having good-quality BOLD data. Using alimited number of components for the ICA decompositionmay then probably result in a better solution. For theabove reasons, tackling DMN splitting issue remains avery methodological challenge.

CONCLUSION

Here, we proposed a user-independent constrained con-nectivity ICA method with three different selection criteriabased on spatial and temporal information permittingDMN component selection at the single-subject level.When applied to healthy subjects, the three criteria (i.e., (i)spatially pattern driven, (ii) based on an automatic mask-ing driven by the fingerprint properties (time domaindominated), and (iii) based on a compromise between spa-tial and temporal properties) selected the same compo-nent. When subsequently applied to a MCS patient with a‘‘functional hemispherectomy,’’ two of three selection crite-ria were able to detect the proper component. This sug-gests that a method with an automatic masking (in whicha growing number of nodes are repeatedly excluded fromthe network until the selected component satisfies theDMN BOLD fingerprint constraints) could be a successfulapproach to select components in highly pathologicalbrains or in conditions where the network can possiblybecome highly disconnected such as in pharmacologicalcoma [Boveroux et al., 2010]. Defining scores that summa-rize both spatial and temporal fingerprint properties offersa good compromise for DMN connectivity studies in clini-cal settings. When applied to a small cohort of VS patients,the presented methodology permitted to isolate the mainsources of connectivity in the DMN regions and, in ourview, ruled out convincing neuronal contributions. Futurestudies should validate these methodologies comparingtheir sensitivity on simulated data of pathological brains.The robustness of the methodology was illustrated here bythe study of two LIS patients (showing important move-

ment yet permitting extraction of near-normal DMN con-nectivity) and of a MCS patient (showing a half-functioningbrain with similar results obtained by rsfMRI and metabolicPET studies). The proposed approach can also be extendedto other networks with or without renouncing to the con-cept of anticorrelation-corrected connectivity in the assess-ment of resting-state fMRI in real-life clinical settings.

ACKNOWLEDGMENTS

SL is Senior Research Associate; CP is Research Associ-ate; AS, MB, and QN are Postdoctoral Fellows; and MABis Research Fellow at the Fonds de la Recherche Scientifi-que (FRS). The authors thank D. Feyers and C. Bastin foracquisition and sharing of the 11 healthy control data.

REFERENCES

Alkire MT, Miller J (2005): General anesthesia and the neural cor-relates of consciousness. Prog Brain Res 150:229–244.

American Congress of Rehabilitation Medicine (1995): Recommen-dations for use of uniform nomenclature pertinent to patientswith severe alterations in consciousness. Arch Phys MedRehabil 76:205–209.

Anand A, Li Y, Wang Y, Wu J, Gao S, Bukhari L, Mathews VP,Kalnin A, Lowe MJ (2005): Activity and connectivity of brainmood regulating circuit in depression: A functional magneticresonance study. Biol Psychiatry 57:1079–1088.

Baars BJ, Ramsoy TZ, Laureys S (2003): Brain, conscious experi-ence and the observing self. Trends Neurosci 26:671–675.

Beckmann CF, DeLuca M, Devlin JT, Smith SM (2005): Investigationsinto resting-state connectivity using independent componentanalysis. Philos Trans R Soc Lond B Biol Sci 360:1001–1013.

Birn RM, Murphy K, Bandettini PA (2008): The effect of respirationvariations on independent component analysis results of restingstate functional connectivity. Hum Brain Mapp 29:740–750.

Biswal B, Yetkin FZ, Haughton VM, Hyde JS (1995): Functionalconnectivity in the motor cortex of resting human brain usingecho-planar MRI. Magn Reson Med 34:537–541.

Boly M, Phillips C, Balteau E, Schnakers C, Degueldre C, MoonenG, Luxen A, Peigneux P, Faymonville ME, Maquet P, LaureysS (2008): Consciousness and cerebral baseline activity fluctua-tions. Hum Brain Mapp 29:868–874.

Boly M, Tshibanda L, Vanhaudenhuyse A, Noirhomme Q,Schnakers C, Ledoux D, Boveroux P, Garweg C, LambermontB, Phillips C, Luxen A, Moonen G, Bassetti C, Maquet P, Lau-reys S (2009): Functional connectivity in the default networkduring resting state is preserved in a vegetative but not in abrain dead patient. Hum Brain Mapp 30:2393–2400.

Born JD, Albert A, Hans P, Bonnal J (1985): Relative prognosticvalue of best motor response and brain stem reflexes inpatients with severe head injury. Neurosurgery 16:595–601.

Boveroux P, Vanhaudenhuyse A, Bruno MA, Noirhomme Q, Lau-wick S, Luxen A, Degueldre C, Plenevaux A, Schnakers C,Phillips C, Brichant JF, Bonhomme V, Maquet P, Greicius MD,Laureys S, Boly M (2010): Breakdown of within- and between-network resting state functional magnetic resonance imagingconnectivity during propofol-induced loss of consciousness.Anesthesiology 113:1038–1053.

r Soddu et al. r

r 16 r

Calhoun VD, Eichele T, Pearlson G (2009): Functional brain net-works in schizophrenia: A review. Front Hum Neurosci 3:17.

Cole DM, Smith SM, Beckmann CF (2010): Advances and pitfallsin the analysis and interpretation of resting-state FMRI data.Front Syst Neurosci 4:8.

Cordes D, Haughton VM, Arfanakis K, Wendt GJ, Turski PA,Moritz CH, et al. (2000): Mapping functionally related regionsof brain with functional connectivity MR imaging. AJNR Am JNeuroradiol 21:1636–1644.

Dai W, Carmichael OT, Lopez OL, Becker JT, Kuller LH, GachHM (2008): Effects of image normalization on the statisticalanalysis of perfusion MRI in elderly brains. J Magn ResonImaging 28:1351–1360.

Damoiseaux JS, Rombouts SA, Barkhof F, Scheltens P, StamCJ, SmithSM, Beckmann CF (2006): Consistent resting-state networksacross healthy subjects. Proc Natl Acad Sci USA 103:13848–13853.

De Luca M, Beckmann CF, De Stefano N, Matthews PM, SmithSM (2006): fMRI resting state networks define distinct modesof long-distance interactions in the human brain. Neuroimage29:1359–1367.

De Martino F, Gentile F, Esposito F, Balsi M, Di Salle F, Goebel R,Formisano E (2007): Classification of fMRI independent com-ponents using IC-fingerprints and support vector machineclassifiers. Neuroimage 34:177–194.

Esposito F, Scarabino T, Hyvarinen A, Himberg J, Formisano E,Comani S, Tedeschi G, Goebel R, Seifritz E, Di Salle F (2005):Independent component analysis of fMRI group studies byself-organizing clustering. Neuroimage 25:193–205.

Esposito F, Aragri A, Pesaresi I, Cirillo S, Tedeschi G, Marciano E,Goebel R, Di Salle F (2008): Independent component model ofthe default-mode brain function: Combining individual-leveland population-level analyses in resting-state fMRI. MagnReson Imaging 26:905–913.

Fair DA, Cohen AL, Dosenbach NU, Church JA, Miezin FM,Barch DM, Raichle ME, Petersen SE, Schlaggar BL (2008): Thematuring architecture of the brain’s default network. Proc NatlAcad Sci USA 105:4028–4032.

Formisano E, Esposito F, Di Salle F, Goebel R (2004): Cortex-basedindependent component analysis of fMRI time series. MagnReson Imaging 22:1493–1504.

Fox MD, Raichle ME (2007): Spontaneous fluctuations in brain ac-tivity observed with functional magnetic resonance imaging.Nat Rev Neurosci 8:700–711.

Fox MD, Snyder AZ, Vincent JL, Corbetta M, Van Essen DC,Raichle ME (2005): The human brain is intrinsically organizedinto dynamic, anticorrelated functional networks. Proc NatlAcad Sci USA 102:9673–9678.

Fox MD, Zhang D, Snyder AZ, Raichle ME (2009): The global sig-nal and observed anticorrelated resting state brain networks. JNeurophysiol 101:3270–3283.

Giacino JT, Kalmar K, Whyte J (2004): The JFK Coma RecoveryScale-Revised: Measurement characteristics and diagnostic util-ity. Arch Phys Med Rehabil 85:2020–2029.

Golland Y, Bentin S, Gelbard H, Benjamini Y, Heller R, Nir Y,Hasson U, Malach R (2007): Extrinsic and intrinsic systems inthe posterior cortex of the human brain revealed during natu-ral sensory stimulation. Cereb Cortex 17:766–777.

Golland Y, Golland P, Bentin S, Malach R (2008): Data-drivenclustering reveals a fundamental subdivision of the humancortex into two global systems. Neuropsychologia 46:540–553.

Gray MA, Minati L, Harrison NA, Gianaros PJ, Napadow V,Critchley HD (2009): Physiological recordings: Basic concepts

and implementation during functional magnetic resonanceimaging. Neuroimage 47:1105–1115.

Greicius MD, Menon V (2004): Default-mode activity during apassive sensory task: Uncoupled from deactivation but impact-ing activation. J Cogn Neurosci 16:1484–1492.

Greicius MD, Krasnow B, Reiss AL, Menon V (2003): Functional con-nectivity in the resting brain: A network analysis of the defaultmode hypothesis. Proc Natl Acad Sci USA 100:253–258.

Greicius MD, Srivastava G, Reiss AL, Menon V (2004): Default-mode network activity distinguishes Alzheimer’s disease fromhealthy aging: Evidence from functional MRI. Proc Natl AcadSci USA 101:4637–4642.

Gusnard DA, Akbudak E, Shulman GL, Raichle ME (2001): Medialprefrontal cortex and self-referential mental activity: Relationto a default mode of brain function. Proc Natl Acad Sci USA98:4259–4264.

Hagmann P, Cammoun L, Gigandet X, Gerhard S, Ellen Grant P,Wedeen V, Meuli R, Thiran JP, Honey CJ, Sporns O (2008): Map-ping the structural core of human cerebral cortex. PLoS Biol 6:e159.

Hyvarinen A, Karhunen J, Oja E (2001): Independent ComponentAnalysis. John Wiley & Sons.

Jafri MJ, Pearlson GD, Stevens M, Calhoun VD (2008): A method forfunctional network connectivity among spatially independent rest-ing-state components in schizophrenia. Neuroimage 39:1666–1681.

Kennedy DP, Redcay E, Courchesne E (2006): Failing to deacti-vate: Resting functional abnormalities in autism. Proc NatlAcad Sci USA 103:8275–8280.

Kiviniemi V, Kantola JH, Jauhiainen J, Hyvarinen A, Tervonen O(2003): Independent component analysis of nondeterministicfMRI signal sources. Neuroimage 19:253–260.

Kiviniemi V, Starck T, Remes J, Long X, Nikkinen J, Haapea M,Veijola J, Moilanen I, Isohanni M, Zang YF, Tervonen O (2009):Functional segmentation of the brain cortex using high modelorder group PICA. Hum Brain Mapp 30:3865–3886.

Koch W, Teipel S, Mueller S, Buerger K, Bokde ALW, Hampel H,Coates U, Reiser M (2010): Effects of aging on default modenetwork activity in resting state fMRI: Does the method ofanalysis matter? Neuroimage 15:280–287.

Laureys S, Owen AM, Schiff ND (2004): Brain function in coma,vegetative state, and related disorders. Lancet Neurol 3:537–546.

Laureys S, Perrin F, Bredart S (2007): Self-consciousness in non-commu-nicative patients. ConsciousCogn 16:722–741; discussion 742–725.

Lowe MJ, Mock BJ, Sorenson JA (1998): Functional connectivity insingle and multislice echoplanar imaging using resting-statefluctuations. Neuroimage 7:119–132.

Lowe MJ, Phillips MD, Lurito JT, Mattson D, Dzemidzic M, Math-ews VP (2002): Multiple sclerosis: Low-frequency temporalblood oxygen level-dependent fluctuations indicate reducedfunctional connectivity initial results. Radiology 224:184–192.

Maquet P (1997): Positron emission tomography studies of sleepand sleep disorders. J Neurol 244:S23–S28.

McKeown MJ, Makeig S, Brown GG, Jung TP, Kindermann SS,Bell AJ, Sejnowski T (1998): Analysis of fMRI data by blindseparation into independent spatial components. Hum BrainMapp 6:160–188.

Meindl T, Teipel S, Elmouden R, Mueller S, Koch W, Dietrich O,Coates U, Reiser M, Glaser C (2010): Test-retest reproducibilityof the default-mode network in healthy individuals. HumBrain Mapp 31: 237–246.

Minoshima S, Giordani B, Berent S, Frey KA, Foster NL, Kuhl DE(1997): Metabolic reduction in the posterior cingulate cortex invery early Alzheimer’s disease. Ann Neurol 42:85–94.


r 17 r

Mitra PP, Ogawa S, Hu X, Ugurbil K (1997): The nature of spatiotem-poral changes in cerebral hemodynamics as manifested in func-tionalmagnetic resonance imaging.MagnResonMed 37:511–518.

Murphy K, Birn RM, Handwerker DA, Jones TB, Bandettini PA(2009): The impact of global signal regression on resting statecorrelations: Are anti-correlated networks introduced? Neuro-image 44:893–905.

Nir Y, Hasson U, Levy I, Yeshurun Y, Malach R (2006): Wide-spread functional connectivity and fMRI fluctuations in humanvisual cortex in the absence of visual stimulation. Neuroimage30:1313–1324.

Nir Y, Mukamel R, Dinstein I, Privman E, Harel M, Fisch L,Gelbard-Sagiv H, Kipervasser S, Andelman F, Neufeld MY,Kramer U, Arieli A, Fried I, Malach R (2008): Interhemisphericcorrelations of slow spontaneous neuronal fluctuationsrevealed in human sensory cortex. Nat Neurosci 11:1100–1108.

Perlbarg V, Marrelec G (2008): Contribution of exploratory meth-ods to the investigation of extended large-scale brain networksin functional MRI: Methodologies, results, and challenges. Int JBiomed Imaging 2008:218519.

Perlbarg V, Bellec P, Anton JL, Pelegrini-Issac M, Doyon J, BenaliH (2007): CORSICA: Correction of structured noise in fMRI byautomatic identification of ICA components. Magn ResonImaging 25:35–46.

Rainville P, Hofbauer RK, Bushnell MC, Duncan GH, Price DD(2002): Hypnosis modulates activity in brain structures involvedin the regulation of consciousness. J Cogn Neurosci 14:887–901.

Rombouts SA, Damoiseaux JS, Goekoop R, Barkhof F, Scheltens P,Smith SM, Beckmann CF (2009): Model-free group analysisshows altered BOLD FMRI networks in dementia. Hum BrainMapp 30:256–266.

Schilbach L, Eickhoff SB, Rotarska-Jagiela A, Fink GR, Vogeley K

(2008): Minds at rest? Social cognition as the default mode of

cognizing and its putative relationship to the ‘‘default system’’

of the brain. Conscious Cogn 17:457–467.

Smith SM, Fox PT, Miller KL, Glahn DC, Fox PM, Mackay CE, Fil-

ippini N, Watkins KE, Toro R, Laird AR, Beckmann CF (2009):

Correspondence of the brain’s functional architecture during

activation and rest. Proc Natl Acad Sci USA 106:13040–13045.Soddu A, Vanhaudenhuyse A, Demertzi A, Bruno MA, Tshibanda

L, Di H, Boly M, Papa M, Laureys S, Noirhomme Q: Restingstate activity in patients with disorders of consciousness. FunctNeurol 2011 (in press).

The Multi-Society Task Force on PVS (1994): Medical aspects of the

persistent vegetative state (1). N Engl J Med 330:1499–1508.Vanhaudenhuyse A, Noirhomme Q, Tshibanda LJ, Bruno MA,

Boveroux P, Schnakers C, Soddu A, Perlbarg V, Ledoux D, Bri-chant JF, Moonen G, Maquet P, Greicius MD, Laureys S, BolyM (2010): Default network connectivity reflects the level ofconsciousness in non-communicative brain-damaged patients.Brain 133:161–171.

Vanhaudenhuyse A, Demertzi A, Schabus M, Noirhomme Q,

Bredart S, Boly M, Phillips C, Soddu A, Luxen A, Moonen G,

Laureys S (2011): Two distinct neuronal networks mediate the

awareness of environment and of self. J Cogn Neurosci

23(3):570–578.

Vincent JL, Patel GH, Fox MD, Snyder AZ, Baker JT, Van Essen

DC, Zempel JM, Snyder LH, Corbetta M, Raichle ME (2007):

Intrinsic functional architecture in the anaesthetized monkey

brain. Nature 447:83–86.

Weissman-Fogel I, Moayedi M, Taylor KS, Pope G, Davis KD

(2010): Cognitive and default-mode resting state networks: Do

male and female brains ‘‘rest’’ differently? Hum Brain Mapp

31:1713–1726.Xiong J, Parsons LM, Gao JH, Fox PT (1999): Interregional connec-

tivity to primary motor cortex revealed using MRI resting stateimages. Hum Brain Mapp 8:151–156.

Yang H, Long XY, Yang Y, Yan H, Zhu CZ, Zhou XP, Zang YF,Gong QY (2007): Amplitude of low frequency fluctuationwithin visual areas revealed by resting-state functional MRI.Neuroimage 36:144–152.

Ylipaavalniemi J, Vigario R (2008): Analyzing consistency of independ-ent components: An fMRI illustration. Neuroimage 39:169–180.

APPENDIX

Connectivity Graph and

Anticorrelation-Corrected Number of Edges

To study connectivity between each pair of targetpoints, we implemented a new method, which is based onICA followed by a GLM. After running ICA with 30 com-ponents, we used the corresponding time courses toregress in the BOLD signal in each of the 13 ROIs (averagetime course over the voxels belonging to the ROI). Foreach component, we obtained 13 parameter estimates (betavalues) indicating the weight of each regressor and thecorresponding T-values:

TCROI1 ¼ b0;1 þ b1;1TCIC1 þ ::: ::: ::: þ b30;1TCIC30 þ e1;1

TCROI2 ¼ b0;2 þ b1;2TCIC1þ ::: ::: ::: þ b30;2TCIC30

þ e1;2

..

. ... ..

. ...

TCROIn ¼ b0;n þ b1;nTCIC1 þ ::: ::: ::: þ b30;nTCIC30 þ e1;n

T ¼ cT � bffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiVarðeÞ � cT � ðXTXÞ�1 � c

q ;

where c is the contrast vector that isolates each IC, and Xis the matrix with the predictors TCIC. The anticorrelation-corrected number of edges was obtained by multiplyingthe total number of edges of each graph by a weight ‘‘w,’’which measures the anticorrelation of the DMN activitywith the extrinsic system/ECN. The anticorrelation-cor-rected number of edges EAntiCC was then defined in termsof the number of edges E as:

EAntiCC ¼ E

2� 1� hTextri

maxðjTextrjÞ

� �¼ E � w;

where hTextri is the average of the five extrinsic ROIs T-values, max(|Textr|) is the maximum of the absolute valueof the five extrinsic ROIs T-values and � signs hold,respectively, for the ICs 1–30 (minus sign) and for ICs 31–60 (plus sign). The anticorrelation index would be close to0 if the activity in the extrinsic network highly correlateswith the activity in the intrinsic network (hTextri/

r Soddu et al. r

r 18 r

max(|Textr|) subtracts to 1), e.g., a global signal, and closeto 1 if the activity in the extrinsic network is anticorrelatedwith the activity in the DMN (hTextri/max(|Textr|) sumsup to 1), e.g., the DMN component.

Fingerprint and

Anticorrelation-Corrected Scores

A fingerprint (as in De Martino et al. [2007], the finger-print reports normalized values, respectively, for: degree

of clustering, skewness, kurtosis, spatial entropy as calcu-lated from the distribution of the z values for the consid-ered IC and one-lag autocorrelation, temporal entropy,and the power of the five frequency bands [0–0.008, 0.008–0.02, 0.02–0.05, 0.05–0.1, and 0.1–0.25 Hz] as extractedfrom the time course of the considered IC) is calculated inBrain Voyager for each IC. As presented in the methods todrive the automatic masking or to build the anticorrela-tion-corrected scores, a Euclidean distance in the finger-print space is defined as:

using an average fingerprint with values ‘‘ : : : neur’’ builtfrom the DMN components of the 11 healthy subjects(Group 1). The ‘‘anticorrelation-corrected score’’ was builtby multiplying the number of anticorrelation-correctededges by a new weight ‘‘wF’’ defined by:

wF ¼ 1� DIC

maxðDICÞ

� �

with wF close to 0 for components that have a dis-

tance comparable with the max(DIC) (‘‘artifacts’’ compo-

nents) and close to 1 for components with ‘‘neuronal’’origin.

Motion Indices

Two motion indices were introduced describing the motionof patients compared with healthy controls. The first index Dmeasures the mean over time of the displacement during thefull acquisition (translation measured in millimeters and rota-tions measured in degrees). The second index R measuresthe mean over time of the displacement speed (variation overa repetition time) during the full acquisition:

DIC ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðClusterIC � ClusterneurÞ2 þ ðSkewIC � SkewneurÞ2 þ ðKurtIC � KurtneurÞ2

q

þðSpatEntrIC � SpatEntrneurÞ2 þ ðOneLagACIC �OneLagACneurÞ

2 þ ðTempEntrIC � TempEntrneurÞ2

þðPð0� 0:008 HzÞIC � Pð0� 0:008 HzÞneurÞ2 þ ðPð0:008� 0:02 HzÞIC � Pð0:008� 0:02 HzÞneurÞ

2

þðPð0:02� 0:05 HzÞIC � Pð0:02� 0:05 HzÞneurÞ2 þ ðPð0:05� 0:1 HzÞIC � Pð0:05� 0:1 HzÞneurÞ

2

þðPð0:1� 0:25 HzÞIC � Pð0:1� 0:25 HzÞneurÞ2

D ¼D ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

TraX2 þ TraY2 þ TraZ2 þ RotX2 þ RotY2 þ RotZ2p E

R ¼D ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

DTRTraX2 þ DTRTraY2 þ DTRTraZ2 þ DTRRotX2 þ DTRRotY2 þ DTRRotZ2p E

;

where DTR indicates the variation of a parameter over a TR.


r 19 r

Documents

Identifying the default-mode component in spatial IC analyses of